Watching Lee Sedol — one of the most skilled humans on the planet when it comes to the ancient game of Go — lose to a machine over and over again is hard to take.
The 34-year-old has spent almost his entire life practicing and perfecting the game of Go. He’s a national treasure in his home country of South Korea and he’s renowned for playing beautiful moves that other players can’t think of.
But his abilities proved to be no match for AlphaGo — an algorithm developed by Google DeepMind, which lends its name to the new film directed by Gary Kohs. Business Insider went to watch the film’s UK premiere at the BFI London Film Festival on Monday.
The artificially intelligent AlphaGo algorithm, programmed to learn how to play Go by analysing previous Go games and playing against older versions of itself, beat Lee Sedol 4-1 in a five match tournament at a Four Seasons hotel in Seoul last March.
Lee Sedol confidently strolled into the tournament expecting to beat AlphaGo 5-0, or 4-1 at the very least. The look on his face after he lost the first game is something to behold. Lee Sedol doesn’t usually lose, so naturally he’s confused and in shock. At this point in the documentary you realise the ability of the AI and you can’t help but feel slightly scared. You’re very much on the side of Lee Sedol, or to put it another way, humanity as a whole.
Lee Sedol adopted new tactics in the second and third games to try to figure out AlphaGo’s weaknesses but it didn’t get him anywhere. If anything it just made it easier for AlphaGo.
During these early games, Lee Sedol can be seen pausing for thought while having a cigarette on the luxury hotel terrace. He looks like a broken man as he tries to understand what’s happening. “I was genuinely tearing up for Lee Sedol after the third match,” one viewer told Business Insider after the screening.
But something special happened in the fourth game: Lee Sedol won. It’s a great moment and it makes you feel like humanity isn’t out of it yet. (“Screw you machines, we’re not going down without a fight.”) AlphaGo goes on to win the fifth game but it doesn’t matter because at least we’ve got one.
The question at the end of the film is: “What next for AlphaGo?” Well, DeepMind has officially retired the algorithm.
Naturally, there were some concerns in the audience when DeepMind CEO Demis Hassabis and AlphaGo lead programmer David Silver appeared on stage with the director and Fan Hui, the current European Go champion and an advisor to DeepMind.
DeepMind CEO Demis Hassabis said it would be wrong to regulate AI today
One member of the audience highlighted that powerful AI has its risks and asked whether it should be regulated in any way.
“With any new powerful technology, and AI is no different, I think the technology itself is neutral,” replied Hassabis. “But it’s how we decide to use it that’s going to determine whether it’s good or bad.
“It’s still very early stage. We won at Go. But the real world is way more complex than even Go is so there’s a long way to go even though we’ve made some impressive advances. That gives us the time in the research community to research all these deep questions about what are the right control mechanisms, what are the right ways to specify goals, and what are the limits of these systems.
“We need a lot more research first then I think we’ll know what are the right protocols and safety mechanisms that should be applied globally across all these technologies and at that point we can start thinking about regulation. But until then, you wouldn’t know what the right things were to regulate.”